tdd-guide
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external content in the form of source code and coverage data processed through
scripts/test_generator.pyandscripts/coverage_analyzer.py. The absence of explicit boundary markers or instructions to isolate analyzed data from the agent's control flow creates an indirect prompt injection surface. \n - Ingestion points: External code and reports (LCOV, JSON, XML) handled by
scripts/coverage_analyzer.pyandscripts/test_generator.py. \n - Boundary markers: None identified; the skill does not employ specific delimiters or guidance to ignore instructions embedded within user-supplied content. \n
- Capability inventory: The skill includes Python scripts for metrics calculation and test scaffolding; no subprocess calls, file system writes, or network operations are present in these tools. \n
- Sanitization: No specific input sanitization is implemented to detect or neutralize adversarial prompts in the ingested data.
Audit Metadata